trained result totally different than the test result

[LouisCortes]

I trained a model giving back this result :

with this comand : python train.py --label_nc 0 --no_instance --name Dal --dataroot .\pix2pixHD\datasets\Dal --save_epoch_freq 5

But when I test the model I get this :

with this command : python test.py --name Dal --netG local --ngf 32 --resize_or_crop none --label_nc 0 --no_instance --dataroot "./pix2pixHD/Test"

I don't know if it's normal, there seems to be a very big difference between the two.

[DuWenTao2]

I also meet this problem , can you solve it?

[DuWenTao2]

hello I have same problem as you , are you sloving it ?

[kizum]

The command you used to train the network only train the global generator. But the test command you used here is supposed to run with the local enhancer + global generator. (the '--netG local' argument)

So modify the test command to:
python test.py --name Dal --ngf 32 --resize_or_crop none --label_nc 0 --no_instance --dataroot "./pix2pixHD/Test
And it should be fine.